aboutsummaryrefslogtreecommitdiffstats
path: root/libavfilter/x86/vf_fspp.asm
blob: c7f8f64f1b460c313528571f72e2101155428640 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
;*****************************************************************************
;* x86-optimized functions for fspp filter
;*
;* Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
;* Copyright (C) 2005 Nikolaj Poroshin <porosh3@psu.ru>
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or modify
;* it under the terms of the GNU General Public License as published by
;* the Free Software Foundation; either version 2 of the License, or
;* (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;* GNU General Public License for more details.
;*
;* You should have received a copy of the GNU General Public License along
;* with FFmpeg; if not, write to the Free Software Foundation, Inc.,
;* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
;******************************************************************************

%include "libavutil/x86/x86util.asm"

SECTION_RODATA

pb_dither: db 0,  48,  12,  60,   3,  51,  15,  63, 32,  16,  44,  28,  35,  19,  47,  31, \
              8,  56,   4,  52,  11,  59,   7,  55, 40,  24,  36,  20,  43,  27,  39,  23, \
              2,  50,  14,  62,   1,  49,  13,  61, 34,  18,  46,  30,  33,  17,  45,  29, \
             10,  58,   6,  54,   9,  57,   5,  53, 42,  26,  38,  22,  41,  25,  37,  21
pw_187E: times 4 dw 0x187E ; FIX64(0.382683433, 14)
pw_22A3: times 4 dw 0x22A3 ; FIX64(1.082392200, 13)
pw_2D41: times 4 dw 0x2D41 ; FIX64(1.414213562, 13)
pw_539F: times 4 dw 0x539F ; FIX64(1.306562965, 14)
pw_5A82: times 4 dw 0x5A82 ; FIX64(1.414213562, 14)
pw_3B21: times 4 dw 0x3B21 ; FIX64(1.847759065, 13)
pw_AC62: times 4 dw 0xAC62 ; FIX64(-2.613125930, 13)
pw_3642: times 4 dw 0x3642 ; FIX64(0.847759065, 14)
pw_2441: times 4 dw 0x2441 ; FIX64(0.566454497, 14)
pw_0CBB: times 4 dw 0x0CBB ; FIX64(0.198912367, 14)
pw_4:    times 4 dw 4
pw_2:    times 4 dw 2

SECTION .text

%define DCTSIZE 8

INIT_MMX mmx

;void ff_store_slice_mmx(uint8_t *dst, int16_t *src,
;                        ptrdiff_t dst_stride, ptrdiff_t src_stride,
;                        ptrdiff_t width, ptrdiff_t height, ptrdiff_t log2_scale)
%if ARCH_X86_64
cglobal store_slice, 7, 9, 0, dst, src, dst_stride, src_stride, width, dither_height, dither, tmp, tmp2
%else
cglobal store_slice, 2, 7, 0, dst, src, width, dither_height, dither, tmp, tmp2
%define dst_strideq r2m
%define src_strideq r3m
    mov       widthq, r4m
    mov       dither_heightq, r5m
    mov       ditherq, r6m ; log2_scale
%endif
    add       widthq, 7
    mov       tmpq, src_strideq
    and       widthq, ~7
    sub       dst_strideq, widthq
    movd      m5, ditherd ; log2_scale
    xor       ditherq, -1 ; log2_scale
    mov       tmp2q, tmpq
    add       ditherq, 7 ; log2_scale
    neg       tmpq
    sub       tmp2q, widthq
    movd      m2, ditherd ; log2_scale
    add       tmp2q, tmp2q
    lea       ditherq, [pb_dither]
    mov       src_strideq, tmp2q
    shl       tmpq, 4
    lea       dither_heightq, [ditherq+dither_heightq*8]
    pxor      m7, m7

.loop_height:
    movq      m3, [ditherq]
    movq      m4, m3
    punpcklbw m3, m7
    punpckhbw m4, m7
    mov       tmp2q, widthq
    psraw     m3, m5
    psraw     m4, m5

.loop_width:
    movq      [srcq+tmpq], m7
    movq      m0, [srcq]
    movq      m1, [srcq+8]
    movq      [srcq+tmpq+8], m7
    paddw     m0, m3
    paddw     m1, m4
    movq      [srcq], m7
    psraw     m0, m2
    psraw     m1, m2
    movq      [srcq+8], m7
    packuswb  m0, m1
    add       srcq, 16
    movq      [dstq], m0
    add       dstq, 8
    sub       tmp2q, 8
    jg .loop_width

    add       srcq, src_strideq
    add       ditherq, 8
    add       dstq, dst_strideq
    cmp       ditherq, dither_heightq
    jl .loop_height
    RET

;void ff_store_slice2_mmx(uint8_t *dst, int16_t *src,
;                         ptrdiff_t dst_stride, ptrdiff_t src_stride,
;                         ptrdiff_t width, ptrdiff_t height, ptrdiff_t log2_scale)
%if ARCH_X86_64
cglobal store_slice2, 7, 9, 0, dst, src, dst_stride, src_stride, width, dither_height, dither, tmp, tmp2
%else
cglobal store_slice2, 0, 7, 0, dst, src, width, dither_height, dither, tmp, tmp2
%define dst_strideq r2m
%define src_strideq r3m
    mov       dstq, dstm
    mov       srcq, srcm
    mov       widthq, r4m
    mov       dither_heightq, r5m
    mov       ditherq, r6m ; log2_scale
%endif
    add       widthq, 7
    mov       tmpq, src_strideq
    and       widthq, ~7
    sub       dst_strideq, widthq
    movd      m5, ditherd ; log2_scale
    xor       ditherq, -1 ; log2_scale
    mov       tmp2q, tmpq
    add       ditherq, 7 ; log2_scale
    sub       tmp2q, widthq
    movd      m2, ditherd ; log2_scale
    add       tmp2q, tmp2q
    lea       ditherq, [pb_dither]
    mov       src_strideq, tmp2q
    shl       tmpq, 5
    lea       dither_heightq, [ditherq+dither_heightq*8]
    pxor      m7, m7

.loop_height:
    movq      m3, [ditherq]
    movq      m4, m3
    punpcklbw m3, m7
    punpckhbw m4, m7
    mov       tmp2q,widthq
    psraw     m3, m5
    psraw     m4, m5

.loop_width:
    movq      m0, [srcq]
    movq      m1, [srcq+8]
    paddw     m0, m3
    paddw     m0, [srcq+tmpq]
    paddw     m1, m4
    movq      m6, [srcq+tmpq+8]
    movq      [srcq+tmpq], m7
    psraw     m0, m2
    paddw     m1, m6
    movq      [srcq+tmpq+8], m7
    psraw     m1, m2
    packuswb  m0, m1
    movq      [dstq], m0
    add       srcq, 16
    add       dstq, 8
    sub       tmp2q, 8
    jg .loop_width

    add       srcq, src_strideq
    add       ditherq, 8
    add       dstq, dst_strideq
    cmp       ditherq, dither_heightq
    jl .loop_height
    RET

;void ff_mul_thrmat_mmx(int16_t *thr_adr_noq, int16_t *thr_adr, int q);
cglobal mul_thrmat, 3, 3, 0, thrn, thr, q
    movd      m7, qd
    movq      m0, [thrnq]
    punpcklwd m7, m7
    movq      m1, [thrnq+8]
    punpckldq m7, m7
    pmullw    m0, m7
    movq      m2, [thrnq+8*2]
    pmullw    m1, m7
    movq      m3, [thrnq+8*3]
    pmullw    m2, m7
    movq      [thrq], m0
    movq      m4, [thrnq+8*4]
    pmullw    m3, m7
    movq      [thrq+8], m1
    movq      m5, [thrnq+8*5]
    pmullw    m4, m7
    movq      [thrq+8*2], m2
    movq      m6, [thrnq+8*6]
    pmullw    m5, m7
    movq      [thrq+8*3], m3
    movq      m0, [thrnq+8*7]
    pmullw    m6, m7
    movq      [thrq+8*4], m4
    movq      m1, [thrnq+8*7+8]
    pmullw    m0, m7
    movq      [thrq+8*5], m5
    movq      m2, [thrnq+8*7+8*2]
    pmullw    m1, m7
    movq      [thrq+8*6], m6
    movq      m3, [thrnq+8*7+8*3]
    pmullw    m2, m7
    movq      [thrq+8*7], m0
    movq      m4, [thrnq+8*7+8*4]
    pmullw    m3, m7
    movq      [thrq+8*7+8], m1
    movq      m5, [thrnq+8*7+8*5]
    pmullw    m4, m7
    movq      [thrq+8*7+8*2], m2
    movq      m6, [thrnq+8*7+8*6]
    pmullw    m5, m7
    movq      [thrq+8*7+8*3], m3
    movq      m0, [thrnq+14*8]
    pmullw    m6, m7
    movq      [thrq+8*7+8*4], m4
    movq      m1, [thrnq+14*8+8]
    pmullw    m0, m7
    movq      [thrq+8*7+8*5], m5
    pmullw    m1, m7
    movq      [thrq+8*7+8*6], m6
    movq      [thrq+14*8], m0
    movq      [thrq+14*8+8], m1
    RET

%macro COLUMN_FDCT 1-3 0, 0
    movq      m1, [srcq+DCTSIZE*0*2]
    movq      m7, [srcq+DCTSIZE*3*2]
    movq      m0, m1
    paddw     m1, [srcq+DCTSIZE*7*2]
    movq      m3, m7
    paddw     m7, [srcq+DCTSIZE*4*2]
    movq      m5, m1
    movq      m6, [srcq+DCTSIZE*1*2]
    psubw     m1, m7
    movq      m2, [srcq+DCTSIZE*2*2]
    movq      m4, m6
    paddw     m6, [srcq+DCTSIZE*6*2]
    paddw     m5, m7
    paddw     m2, [srcq+DCTSIZE*5*2]
    movq      m7, m6
    paddw     m6, m2
    psubw     m7, m2
    movq      m2, m5
    paddw     m5, m6
    psubw     m2, m6
    paddw     m7, m1
    movq      m6, [thrq+4*16+%2]
    psllw     m7, 2
    psubw     m5, [thrq+%2]
    psubw     m2, m6
    paddusw   m5, [thrq+%2]
    paddusw   m2, m6
    pmulhw    m7, [pw_2D41]
    paddw     m5, [thrq+%2]
    paddw     m2, m6
    psubusw   m5, [thrq+%2]
    psubusw   m2, m6
    paddw     m5, [pw_2]
    movq      m6, m2
    paddw     m2, m5
    psubw     m5, m6
    movq      m6, m1
    paddw     m1, m7
    psubw     m1, [thrq+2*16+%2]
    psubw     m6, m7
    movq      m7, [thrq+6*16+%2]
    psraw     m5, 2
    paddusw   m1, [thrq+2*16+%2]
    psubw     m6, m7
    paddw     m1, [thrq+2*16+%2]
    paddusw   m6, m7
    psubusw   m1, [thrq+2*16+%2]
    paddw     m6, m7
    psubw     m3, [srcq+DCTSIZE*4*2]
    psubusw   m6, m7
    movq      m7, m1
    psraw     m2, 2
    psubw     m4, [srcq+DCTSIZE*6*2]
    psubw     m1, m6
    psubw     m0, [srcq+DCTSIZE*7*2]
    paddw     m6, m7
    psraw     m6, 2
    movq      m7, m2
    pmulhw    m1, [pw_5A82]
    paddw     m2, m6
    movq      [rsp], m2
    psubw     m7, m6
    movq      m2, [srcq+DCTSIZE*2*2]
    psubw     m1, m6
    psubw     m2, [srcq+DCTSIZE*5*2]
    movq      m6, m5
    movq      [rsp+8*3], m7
    paddw     m3, m2
    paddw     m2, m4
    paddw     m4, m0
    movq      m7, m3
    psubw     m3, m4
    psllw     m3, 2
    psllw     m7, 2
    pmulhw    m3, [pw_187E]
    psllw     m4, 2
    pmulhw    m7, [pw_22A3]
    psllw     m2, 2
    pmulhw    m4, [pw_539F]
    paddw     m5, m1
    pmulhw    m2, [pw_2D41]
    psubw     m6, m1
    paddw     m7, m3
    movq      [rsp+8], m5
    paddw     m4, m3
    movq      m3, [thrq+3*16+%2]
    movq      m1, m0
    movq      [rsp+8*2], m6
    psubw     m1, m2
    paddw     m0, m2
    movq      m5, m1
    movq      m2, [thrq+5*16+%2]
    psubw     m1, m7
    paddw     m5, m7
    psubw     m1, m3
    movq      m7, [thrq+16+%2]
    psubw     m5, m2
    movq      m6, m0
    paddw     m0, m4
    paddusw   m1, m3
    psubw     m6, m4
    movq      m4, [thrq+7*16+%2]
    psubw     m0, m7
    psubw     m6, m4
    paddusw   m5, m2
    paddusw   m6, m4
    paddw     m1, m3
    paddw     m5, m2
    paddw     m6, m4
    psubusw   m1, m3
    psubusw   m5, m2
    psubusw   m6, m4
    movq      m4, m1
    por       m4, m5
    paddusw   m0, m7
    por       m4, m6
    paddw     m0, m7
    packssdw  m4, m4
    psubusw   m0, m7
    movd      tmpd, m4
    or        tmpd, tmpd
    jnz %1
    movq      m4, [rsp]
    movq      m1, m0
    pmulhw    m0, [pw_3642]
    movq      m2, m1
    movq      m5, [outq+DCTSIZE*0*2]
    movq      m3, m2
    pmulhw    m1, [pw_2441]
    paddw     m5, m4
    movq      m6, [rsp+8]
    psraw     m3, 2
    pmulhw    m2, [pw_0CBB]
    psubw     m4, m3
    movq      m7, [outq+DCTSIZE*1*2]
    paddw     m5, m3
    movq      [outq+DCTSIZE*7*2], m4
    paddw     m7, m6
    movq      m3, [rsp+8*2]
    psubw     m6, m0
    movq      m4, [outq+DCTSIZE*2*2]
    paddw     m7, m0
    movq      [outq], m5
    paddw     m4, m3
    movq      [outq+DCTSIZE*6*2], m6
    psubw     m3, m1
    movq      m5, [outq+DCTSIZE*5*2]
    paddw     m4, m1
    movq      m6, [outq+DCTSIZE*3*2]
    paddw     m5, m3
    movq      m0, [rsp+8*3]
    add       srcq, 8+%3
    movq      [outq+DCTSIZE*1*2], m7
    paddw     m6, m0
    movq      [outq+DCTSIZE*2*2], m4
    psubw     m0, m2
    movq      m7, [outq+DCTSIZE*4*2]
    paddw     m6, m2
    movq      [outq+DCTSIZE*5*2], m5
    paddw     m7, m0
    movq      [outq+DCTSIZE*3*2], m6
    movq      [outq+DCTSIZE*4*2], m7
    add       outq, 8+%3
%endmacro

%macro COLUMN_IDCT 0-1 0
    movq      m3, m5
    psubw     m5, m1
    psllw     m5, 1
    paddw     m3, m1
    movq      m2, m0
    psubw     m0, m6
    movq      m1, m5
    psllw     m0, 1
    pmulhw    m1, [pw_AC62]
    paddw     m5, m0
    pmulhw    m5, [pw_3B21]
    paddw     m2, m6
    pmulhw    m0, [pw_22A3]
    movq      m7, m2
    movq      m4, [rsp]
    psubw     m2, m3
    psllw     m2, 1
    paddw     m7, m3
    pmulhw    m2, [pw_2D41]
    movq      m6, m4
    psraw     m7, 2
    paddw     m4, [outq]
    psubw     m6, m7
    movq      m3, [rsp+8]
    paddw     m4, m7
    movq      [outq+DCTSIZE*7*2], m6
    paddw     m1, m5
    movq      [outq], m4
    psubw     m1, m7
    movq      m7, [rsp+8*2]
    psubw     m0, m5
    movq      m6, [rsp+8*3]
    movq      m5, m3
    paddw     m3, [outq+DCTSIZE*1*2]
    psubw     m5, m1
    psubw     m2, m1
    paddw     m3, m1
    movq      [outq+DCTSIZE*6*2], m5
    movq      m4, m7
    paddw     m7, [outq+DCTSIZE*2*2]
    psubw     m4, m2
    paddw     m4, [outq+DCTSIZE*5*2]
    paddw     m7, m2
    movq      [outq+DCTSIZE*1*2], m3
    paddw     m0, m2
    movq      [outq+DCTSIZE*2*2], m7
    movq      m1, m6
    paddw     m6, [outq+DCTSIZE*4*2]
    psubw     m1, m0
    paddw     m1, [outq+DCTSIZE*3*2]
    paddw     m6, m0
    movq      [outq+DCTSIZE*5*2], m4
    add       srcq, 8+%1
    movq      [outq+DCTSIZE*4*2], m6
    movq      [outq+DCTSIZE*3*2], m1
    add       outq, 8+%1
%endmacro

;void ff_column_fidct_mmx(int16_t *thr_adr, int16_t *data, int16_t *output, int cnt);
cglobal column_fidct, 4, 5, 0, 32, thr, src, out, cnt, tmp
.fdct1:
    COLUMN_FDCT .idct1
    jmp .fdct2

.idct1:
    COLUMN_IDCT

.fdct2:
    COLUMN_FDCT .idct2, 8, 16
    sub    cntd, 2
    jg .fdct1
    RET

.idct2:
    COLUMN_IDCT 16
    sub    cntd, 2
    jg .fdct1
    RET

;void ff_row_idct_mmx(int16_t *workspace, int16_t *output_adr, ptrdiff_t output_stride, int cnt);
cglobal row_idct, 4, 5, 0, 16, src, dst, stride, cnt, stride3
    add       strideq, strideq
    lea       stride3q, [strideq+strideq*2]
.loop:
    movq      m0, [srcq+DCTSIZE*0*2]
    movq      m1, [srcq+DCTSIZE*1*2]
    movq      m4, m0
    movq      m2, [srcq+DCTSIZE*2*2]
    punpcklwd m0, m1
    movq      m3, [srcq+DCTSIZE*3*2]
    punpckhwd m4, m1
    movq      m7, m2
    punpcklwd m2, m3
    movq      m6, m0
    punpckldq m0, m2
    punpckhdq m6, m2
    movq      m5, m0
    punpckhwd m7, m3
    psubw     m0, m6
    pmulhw    m0, [pw_5A82]
    movq      m2, m4
    punpckldq m4, m7
    paddw     m5, m6
    punpckhdq m2, m7
    movq      m1, m4
    psllw     m0, 2
    paddw     m4, m2
    movq      m3, [srcq+DCTSIZE*0*2+8]
    psubw     m1, m2
    movq      m2, [srcq+DCTSIZE*1*2+8]
    psubw     m0, m5
    movq      m6, m4
    paddw     m4, m5
    psubw     m6, m5
    movq      m7, m1
    movq      m5, [srcq+DCTSIZE*2*2+8]
    paddw     m1, m0
    movq      [rsp], m4
    movq      m4, m3
    movq      [rsp+8], m6
    punpcklwd m3, m2
    movq      m6, [srcq+DCTSIZE*3*2+8]
    punpckhwd m4, m2
    movq      m2, m5
    punpcklwd m5, m6
    psubw     m7, m0
    punpckhwd m2, m6
    movq      m0, m3
    punpckldq m3, m5
    punpckhdq m0, m5
    movq      m5, m4
    movq      m6, m3
    punpckldq m4, m2
    psubw     m3, m0
    punpckhdq m5, m2
    paddw     m6, m0
    movq      m2, m4
    movq      m0, m3
    psubw     m4, m5
    pmulhw    m0, [pw_AC62]
    paddw     m3, m4
    pmulhw    m3, [pw_3B21]
    paddw     m2, m5
    pmulhw    m4, [pw_22A3]
    movq      m5, m2
    psubw     m2, m6
    paddw     m5, m6
    pmulhw    m2, [pw_2D41]
    paddw     m0, m3
    psllw     m0, 3
    psubw     m4, m3
    movq      m6, [rsp]
    movq      m3, m1
    psllw     m4, 3
    psubw     m0, m5
    psllw     m2, 3
    paddw     m1, m0
    psubw     m2, m0
    psubw     m3, m0
    paddw     m4, m2
    movq      m0, m7
    paddw     m7, m2
    psubw     m0, m2
    movq      m2, [pw_4]
    psubw     m6, m5
    paddw     m5, [rsp]
    paddw     m1, m2
    paddw     m5, m2
    psraw     m1, 3
    paddw     m7, m2
    psraw     m5, 3
    paddw     m5, [dstq]
    psraw     m7, 3
    paddw     m1, [dstq+strideq*1]
    paddw     m0, m2
    paddw     m7, [dstq+strideq*2]
    paddw     m3, m2
    movq      [dstq], m5
    paddw     m6, m2
    movq      [dstq+strideq*1], m1
    psraw     m0, 3
    movq      [dstq+strideq*2], m7
    add       dstq, stride3q
    movq      m5, [rsp+8]
    psraw     m3, 3
    paddw     m0, [dstq+strideq*2]
    psubw     m5, m4
    paddw     m3, [dstq+stride3q*1]
    psraw     m6, 3
    paddw     m4, [rsp+8]
    paddw     m5, m2
    paddw     m6, [dstq+strideq*4]
    paddw     m4, m2
    movq      [dstq+strideq*2], m0
    psraw     m5, 3
    paddw     m5, [dstq]
    psraw     m4, 3
    paddw     m4, [dstq+strideq*1]
    add       srcq, DCTSIZE*2*4
    movq      [dstq+stride3q*1], m3
    movq      [dstq+strideq*4], m6
    movq      [dstq], m5
    movq      [dstq+strideq*1], m4
    sub       dstq, stride3q
    add       dstq, 8
    dec       r3d
    jnz .loop
    RET

;void ff_row_fdct_mmx(int16_t *data, const uint8_t *pixels, ptrdiff_t line_size, int cnt);
cglobal row_fdct, 4, 5, 0, 16, src, pix, stride, cnt, stride3
    lea       stride3q, [strideq+strideq*2]
.loop:
    movd      m0, [pixq]
    pxor      m7, m7
    movd      m1, [pixq+strideq*1]
    punpcklbw m0, m7
    movd      m2, [pixq+strideq*2]
    punpcklbw m1, m7
    punpcklbw m2, m7
    add       pixq,stride3q
    movq      m5, m0
    movd      m3, [pixq+strideq*4]
    movq      m6, m1
    movd      m4, [pixq+stride3q*1]
    punpcklbw m3, m7
    psubw     m5, m3
    punpcklbw m4, m7
    paddw     m0, m3
    psubw     m6, m4
    movd      m3, [pixq+strideq*2]
    paddw     m1, m4
    movq      [rsp], m5
    punpcklbw m3, m7
    movq      [rsp+8], m6
    movq      m4, m2
    movd      m5, [pixq]
    paddw     m2, m3
    movd      m6, [pixq+strideq*1]
    punpcklbw m5, m7
    psubw     m4, m3
    punpcklbw m6, m7
    movq      m3, m5
    paddw     m5, m6
    psubw     m3, m6
    movq      m6, m0
    movq      m7, m1
    psubw     m0, m5
    psubw     m1, m2
    paddw     m7, m2
    paddw     m1, m0
    movq      m2, m7
    psllw     m1, 2
    paddw     m6, m5
    pmulhw    m1, [pw_2D41]
    paddw     m7, m6
    psubw     m6, m2
    movq      m5, m0
    movq      m2, m7
    punpcklwd m7, m6
    paddw     m0, m1
    punpckhwd m2, m6
    psubw     m5, m1
    movq      m6, m0
    movq      m1, [rsp+8]
    punpcklwd m0, m5
    punpckhwd m6, m5
    movq      m5, m0
    punpckldq m0, m7
    paddw     m3, m4
    punpckhdq m5, m7
    movq      m7, m6
    movq      [srcq+DCTSIZE*0*2], m0
    punpckldq m6, m2
    movq      [srcq+DCTSIZE*1*2], m5
    punpckhdq m7, m2
    movq      [srcq+DCTSIZE*2*2], m6
    paddw     m4, m1
    movq      [srcq+DCTSIZE*3*2], m7
    psllw     m3, 2
    movq      m2, [rsp]
    psllw     m4, 2
    pmulhw    m4, [pw_2D41]
    paddw     m1, m2
    psllw     m1, 2
    movq      m0, m3
    pmulhw    m0, [pw_22A3]
    psubw     m3, m1
    pmulhw    m3, [pw_187E]
    movq      m5, m2
    pmulhw    m1, [pw_539F]
    psubw     m2, m4
    paddw     m5, m4
    movq      m6, m2
    paddw     m0, m3
    movq      m7, m5
    paddw     m2, m0
    psubw     m6, m0
    movq      m4, m2
    paddw     m1, m3
    punpcklwd m2, m6
    paddw     m5, m1
    punpckhwd m4, m6
    psubw     m7, m1
    movq      m6, m5
    punpcklwd m5, m7
    punpckhwd m6, m7
    movq      m7, m2
    punpckldq m2, m5
    sub       pixq, stride3q
    punpckhdq m7, m5
    movq      m5, m4
    movq      [srcq+DCTSIZE*0*2+8], m2
    punpckldq m4, m6
    movq      [srcq+DCTSIZE*1*2+8], m7
    punpckhdq m5, m6
    movq      [srcq+DCTSIZE*2*2+8], m4
    add       pixq, 4
    movq      [srcq+DCTSIZE*3*2+8], m5
    add       srcq, DCTSIZE*4*2
    dec       cntd
    jnz .loop
    RET